Group 01 project: Analysis of Gene Expression in Parkinsson Disease

Introduction

Gene expression data (Affymetrix platform).

Collected from the blood of: - 20 healthy patients - 40 patients diagnosed with sporadic Parkinsons disease

Methods

Cleaning and Filtering Dataset

Augmentation

• Mean expression of sets of potential biomarkers
  • Neurodegenerative-associated genes
  • Pro-Inflammation genes

# inflammation index
inflam_list <- c("IL1rn", "TNF", "Saa3", "Emr1", "Adam8", "Itgam")

present_genes <- find_genes(inflam_list, all_genes)
df_wide$inflam_mean <- rowMeans(df_wide[present_genes], na.rm = TRUE)

Analysis

• PCA
  • Normalization
  • PCA calculation
  • Plotting
• Differential Expression Analysis
  • Caclculation of log2 of the expression
  • Fitting linear model to predict the log2 expression based on the condition
  • adjusts p-values -plotting
• Differential Expression Analysis of known biomarker genes
  • Visualization of the mean of biomarker specific gene expression
• KEGG Enrichment Analysis
  • Separates up and down regulated genes
  • Perform KEGG analysis on up or down regulated genes

Results - PCA

Results - Diff. Expression

Results - Diff. Expression of known biomarkers

Neurodegenerative-associated genes

Statistical Significant difference

# Calculate p-value
t_test_result <- t.test(ndg_mean ~ Condition, data = df)

p-value = 0.3512 ## Results - Diff. Expression of known biomarkers

Pro-inflammatory genes

Statistical significant difference

# Calculate p-value
t_test_result <- t.test(inflam_mean ~ Condition, data = df)

p-value = 0.348 ## Kegg enrichment

Kegg pathway

Conclusion

• No clustering / Separtion along the PC’s
• No significant differentially expressed genes
• What to do next?
  • With patient metadata, clustering might present itself
  • Non-linear tendencies. Performing UMAP or t-SNE could provide clustering
  • ANN, KNN logistic modelling could reveal expression patterns
  • Redo the experiment
  • Perhaps blood transcriptomics on PD patient, isn’t the way to go, though biomarker discovery would be ideal.